Presentation Title

Incorporating Model Selection in Predicting The Spread of Invasive Fungal Pathogen Batrachochytrium salamandrivorans

Start Date

November 2016

End Date

November 2016

Location

HUB 302-167

Type of Presentation

Poster

Abstract

Batrachochytrium salamandrivorans (Bsal), a newly emergent and invasive amphibian chytrid fungus, was discovered in 2013 in Western Europe where it has been devastating salamander populations. Species distribution models (SDMs) have been increasingly used to predict and plan for the spread of invasive species. However, few studies statistically evaluate the predictive ability of the models, especially in comparison to alternative models. Here, we use model selection to choose among alternative models for the potential spread to non-native regions of Bsal. Alternative models of spread that have been proposed include climate data, salamander species richness, and salamander import data from the global pet trade. As Bsal has recently been discovered in Europe, we project these models into Europe and use presence and absence data to statistically evaluate the ability of these alternative models to predict Bsal spread. While both AUC and Kappa scores did not provide definitive results, AIC model selection identified a best fit model for the probability of invasion for Bsal - the product of salamander species richness and climate. Our results highlight the difficulty in evaluating success of predictive models when presence data are still limited, such as in the beginning stages of an invasion, but show that these challenges can be overcome with a model selection approach.

This document is currently not available here.

Share

COinS
 
Nov 12th, 1:00 PM Nov 12th, 2:00 PM

Incorporating Model Selection in Predicting The Spread of Invasive Fungal Pathogen Batrachochytrium salamandrivorans

HUB 302-167

Batrachochytrium salamandrivorans (Bsal), a newly emergent and invasive amphibian chytrid fungus, was discovered in 2013 in Western Europe where it has been devastating salamander populations. Species distribution models (SDMs) have been increasingly used to predict and plan for the spread of invasive species. However, few studies statistically evaluate the predictive ability of the models, especially in comparison to alternative models. Here, we use model selection to choose among alternative models for the potential spread to non-native regions of Bsal. Alternative models of spread that have been proposed include climate data, salamander species richness, and salamander import data from the global pet trade. As Bsal has recently been discovered in Europe, we project these models into Europe and use presence and absence data to statistically evaluate the ability of these alternative models to predict Bsal spread. While both AUC and Kappa scores did not provide definitive results, AIC model selection identified a best fit model for the probability of invasion for Bsal - the product of salamander species richness and climate. Our results highlight the difficulty in evaluating success of predictive models when presence data are still limited, such as in the beginning stages of an invasion, but show that these challenges can be overcome with a model selection approach.